1st International ICST Conference on Performance Evaluation Methodologies and Tools

Research Article

Optimizing an OBS scheduler buffer

  • @INPROCEEDINGS{10.1145/1190095.1190123,
        author={Andrew  Zalesky},
        title={Optimizing an OBS scheduler buffer},
        proceedings={1st International ICST Conference on Performance Evaluation Methodologies and Tools},
        publisher={ACM},
        proceedings_a={VALUETOOLS},
        year={2012},
        month={4},
        keywords={Optical burst switching scheduler all-optical switching stochastic performance modeling blocking probability.},
        doi={10.1145/1190095.1190123}
    }
    
  • Andrew Zalesky
    Year: 2012
    Optimizing an OBS scheduler buffer
    VALUETOOLS
    ACM
    DOI: 10.1145/1190095.1190123
Andrew Zalesky1,*
  • 1: Centre for Ultra-Broadband Information Networks (CUBIN), Department of Electrical and Electronic Engineering, The University of Melbourne, Melbourne, VIC 3010, Australia.
*Contact email: a.zalesky@ee.unimelb.edu.au

Abstract

We dimension a buffer in which headers are enqueued during times an OBS scheduler is overwhelmed. Too large a buffer requires a prolonged offset period to counterbalance longer per-hop worst-case header queueing delays, resulting in reduced throughput and increased packet loss due to edge buffer overflow. While too small a buffer cannot absorb enough variability in the header arrival process, resulting in increased burst lost due to a greater number of headers arriving to find a full buffer. For the first time, we develop and analyze a unified OBS model comprising a set of homogeneous and independent edge buffers that feed bursts to a stand-alone link and their headers to a scheduler. Packets arrive at each edge buffer and are assembled into bursts, after which they are transmitted on a wavelength channel if their corresponding headers are successfully processed. To enable a tractable analysis, we invoke an independence and Poisson assumption that permits decoupling of our model into its three constituent sub-models. Simulation is used to gauge the error incurred in invoking these assumptions. We demonstrate that an optimal buffer size may exist and depends on the number of packets comprising a burst and the size of an edge buffer.